I'm building Ditto — it clones websites with 100% visual accuracy and outputs a proper React app with named components and preserved structure.
The problem: you find a design you love, want to use it as a starting point, and your options are either manually recreating it or using a tool that spits out a tangled mess of divs and inline styles. CatchDitto gives you an actual codebase — clean component hierarchy, sensible naming, structure you can extend without wanting to rewrite everything first.
I'm still iterating, would love to hear what others think.
Reading through all these comments gives me hope. Finally people are speaking up about what we have all been experiencing in isolation — the gradual enshittification of Vercel/Next.js in service of Vercel's business model.
The middleware issues the OP describes are not bugs, they are symptoms. When you can't get basic logging to work, when you can't use WebSockets, when you are forced to deploy on edge runtimes that don't support Node.js APIs — these are all deliberate choices to push you toward Vercel's platform. Guillermo and his team have prioritized their cap table over the open web.
This is why I started https://github.com/openuiai/next.js last month. The web is humanity's greatest collaborative achievement. We are building toward a global, decentralized network that should enable unrestricted computation and storage for everyone. And for that to actually happen, we need frameworks with the power and flexibility that Next.js used to have. But instead we are all watching Vercel carve it up to lock developers into their platform
Next.js used to be great. Now it's a trojan horse for vendor lock-in. The fact that so many developers are comparing it to SharePoint and Lotus Notes should be a wake-up call.
To everyone frustrated: you are not wrong. The framework is ACTIVELY working against you unless you are on Vercel.
I'm working on OpenNext.js every single day — still climbing the learning curve of the massive spaghetti codebase, but fully committed. We are already running OpenUI on a fork with full Node.js middleware support and native WebSockets. These will merge into OpenNext.js once I have properly cleaned out the cruft (their CI/CD alone has cost me $400 just figuring out how to untangle it). But I'm taking the time to do this right.
Guillermo: Vercel/Next.js shouldn't dictate where we can deploy. It should respect the open nature of the web. You know this, but you chose the money instead.
I'm building Ditto — it clones websites with 100% visual accuracy and outputs a proper React app with named components and preserved structure.
The problem: you find a design you love, want to use it as a starting point, and your options are either manually recreating it or using a tool that spits out a tangled mess of divs and inline styles. CatchDitto gives you an actual codebase — clean component hierarchy, sensible naming, structure you can extend without wanting to rewrite everything first.
I'm still iterating, would love to hear what others think.